Critical linguistic awareness in the intersection between human language and machine generated linguistic output
While language is inherently human, machines producing linguistic output, like chatbots building on large language models, are now being implemented in various contexts with the intention of providing end users with access to a multitude of content and services, such as modelling texts (e.g., for job applications and school essays), improvement of their personal human language (e.g., revision of texts), and generated texts and translations for humans’ use in their daily activities. In addition, texts partially or fully based on machine generated output are increasingly gaining way into society with or without being marked as such. Machine generated linguistic content is thus becoming a part of people’s daily lives. New technologies can support practices of inclusion and democratization, among other through the linguistic possibilities they afford, such as assistance with translation or captioning (e.g., Zhu et al., 2024). At the same time, digitalization and the increase of new practices in the intersection between human and machine generated language, give rise to new concerns in education and communities, to which educational researchers must respond.
Machine generated linguistic content has characteristics different than those of natural human language, both when it comes to grammar, linguistic variation and lexicon, information structure and hierarchy, social action formation and ascription in human computer interaction and knowledge negotiation in human computer interaction (e.g., Herbold et al. 2023, Durward & Thomson, 2024, Stokoe et al. 2024, Wilson et al., 2024). Language practices such as they emerge in interaction and in texts represent implicit norms, that is norms that people orient to without necessarily being aware of them. This means that, without critical awareness, linguistic patterns from machine generated text may find their way into human language. In addition, machine generated text may involve the presentation of false information (e.g., Chrysostomou et al., 2024, Vindenes, 2024; Hansen, 2024). Interaction with chatbots and other user interfaces building on large language models challenge the end users’ possibilities to critically assess propositions, because the content proposed is fractured from the sources that the presented propositions originated from. As such, the increase of machine generated content in society increases the need for knowledge about the characteristics of the machine generated linguistic output as well as of human language, in addition to the skills and resources required to understand and deal with machine generated linguistic structures and content.
This course gives a broad introduction to language and linguistic practices and uses this as a basis to understand practices emerging in the intersection between human and machine generated language. In addition to perspectives on human language, social interaction and literacy, we give insights into characteristics of machine generated language, social considerations and ideological concerns. Through lectures and discussions, we highlight and discuss differences between human and artificially generated language. We discuss ethical aspects of introducing llm based chatbots in education, especially with consideration to language, including both possibilities and challenges.
The course invites participants researching topics relating to language and linguistic practices and the emergence of machine generated texts with background from various research areas and disciplines, and is especially relevant for students researching different aspects of language and chatbots (llm), including in educational settings,
The course will give insights into fundamentals of language, social interaction and literacy and builds on this to further the discussion about the implementation of machine generated linguistic output in various settings.
Course design and activity:
The course will be held as a 5-day course in June 2025. Participants are invited to present data in a data session or their project if they have not yet acquired data in workshops during the duration of the course. The attendees are expected to submit an essay relating the content of the course to their research topic. The course instructors will provide feedback on the essays. In addition, the course work includes an advanced reading list including papers addressing the basic linguistic content of the course and readings addressing the intersections between human language and machine generated language. TBA: Course program and reading list.
Follow the link to sign up for the course: Summer School 2025 - NORA - Norwegian Artificial Intelligence Research Consortium
Learning outcomes:
On completion of the course, the PhD candidate
Knowledge:
- will have insights about human language and theoretical approaches to human language as well as the intersection between human language and machine generated language
- will have insights into central dilemmas in the intersection between human language and machine generated language
- will have insights into possible consequences of the broad implementation of machine generated linguistic output in society
Skills:
- can assess and critically analyze machine generated linguistic output when it comes to one or more of the course’s main topics (linguistics, social interaction, literacy)
- can assess and critically analyze ways in which machine generated linguistic output is implemented or used in human activities
- Can assess ethical considerations when dealing with machine generated linguistic output in linguistic activity
Competence:
- will have acquired a basis for contributing to the field
- will know and be able to critically discuss the state-of the-art on the intersection between human language and machine generated linguistic output
References:
Chrysostomou, G. Zhao, Z., Williams, M., Aletras, N. (2024). Investigating Hallucinations in Pruned Large Language Models for Abstractive Summarization. Transactions of the Association for Computational Linguistics 2024; 12 1163–1181. doi: https://doi.org/10.1162/tacl_a_00695
Durward, M. & Thomson, C.. 2024. Evaluating Vocabulary Usage in LLMs. In Proceedings of the 19th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2024), pages 266–282, Mexico City, Mexico. Association for Computational Linguistics.
Herbold, S., Hautli-Janisz, A., Heuer, U. et al. A large-scale comparison of human-written versus ChatGPT-generated essays. Sci Rep 13, 18617 (2023). https://doi.org/10.1038/s41598-023-45644-9
Hansen, J. P. B. (2024). Å synliggjøre arbeid. Problemstillinger knyttet til bruken av kunstig intelligens i forskning og undervisning. Academic lecture. Faculty Seminar at the Faculty of teacher education and language, Østfold University College.
Stokoe, E., Albert, S., Buschmeier, H., Strommel, W. (2024). Conversation analysis and conversational technologies: Finding the common ground between academia and industry. Discourse & communication 0(0). Online first. https://doi.org/10.1177/17504813241267118
Vindenes, U. (2024). KI-tekstroboter bør ikke tillates som hjelpemiddel ved eksamen. Opt. ed. Khrono. https://www.khrono.no/ki-tekstroboter-bor-ikke-tillates-som-hjelpemiddel-ved-eksamen/849246
Wilson, M., Petty, J., Frank, R. (2023). How Abstract Is Linguistic Generalization in Large Language Models? Experiments with Argument Structure. Transactions of the Association for Computational Linguistics 2023; 11 1377–1395. doi: https://doi.org/10.1162/tacl_a_00608
Zhu, H., Dai, D.W., Brandt, A., Chen, G., Ferri, G., Hazel, S., Jenks, C., Jones, R., O’Regan, J. & Suzuki, S. Exploring AI for intercultural communication: open conversation. Applied Linguistics Review, 2024. https://doi.org/10.1515/applirev-2024-0186